Skip to main content

Table 3 Comparison of performances (precisions) of Random Forest (RF) and k-Nearest Neighbors (k-NN) classifiers with performances of other methods in published works ([75, 76]) over the same datasets. SM is the single model method; AVG considers the average of the likelihood scores given by the models inferred from five different perturbation random seeds; ∩x/5 considers those predictions from x out of the five models. For RF and k-NN evaluations, probability of perturbation p and likelihood threshold ρ were set to their respective default values p=10% and ρ=0.8

From: Gene function finding through cross-organism ensemble learning

Classifier/Work Method Bostaurus Gallusgallus
RF SM 0.874 0.721
  AVG 0.947 0.833
 1/5 0.794 0.617
 2/5 0.925 0.682
 3/5 0.960 0.857
 4/5 0.967 0.895
 5/5 0.946 0.917
k-NN SM 0.657 0.534
  AVG 0.568 0.690
 1/5 0.625 0.416
 2/5 0.710 0.564
 3/5 0.726 0.742
 4/5 0.656 0.765
 5/5 0.462 0.909
[75] LSI 0.260 -
  LSI-NTN 0.248 -
  LSI-NTM 0.192 -
  LSI-ATN 0.282 -
  SIM 0.190 -
  SIM-NTN 0.206 -
  SIM-NTM 0.240 -
  SIM-ATN 0.322 -
  pLSA 0.206 -
  pLSA-NTN 0.212 -
  pLSA-NTM 0.202 -
  pLSA-ATN 0.162 -
[76] tSVD (LSI) 0.210 0.097
  SIM1 (SIM) 0.157 0.103
  SIM2 0.197 0.097
  pLSA 0.277 0.233
  LDA 0.217 0.127
  AE 0.397 0.397
\